Stevens Team Makes Final Five in Amazon Alexa SocialBot Challenge
As part of the Alexa Prize SocialBot Grand Challenge 5 (SGC5), a team of Stevens Institute of Technology graduate and Ph.D. students has found success with a socialbot that aims to develop a most human concept: friendship.
NAM, or Never Alone with Me, was recently announced as one of five socialbot finalists in the university challenge competition, which focuses on creating an Alexa skill that easily and clearly chats with users on trending topics and news for 20 minutes. The first team to meet the Grand Challenge can win up to a $1 million research grant for their university, along with a science and innovation prize — and Stevens’ NAM team is in the top five.
Launched in 2016, the Alexa Prize SocialBot Grand Challenge is a unique industry-academia partnership program that provides an agile real-world experimentation framework and tools for accelerating scientific discovery. University students can launch innovations online and rapidly adapt their novel technology based on feedback from Alexa customers.
While winning would be a great honor, Jia Xu, the team’s faculty advisor and an assistant professor, says seizing the opportunity to create a fascinating bot that millions of users would like to use in everyday life for chatting, gaining knowledge and entertainment is the primary reason for participating.
The inspiration for NAM comes from the idea that a socialbot can behave like a friend, offering knowledge, helping with problems, caring and sharing personal experiences.
NAM progressed from the previous round of nine finalists based on, among other things, customer feedback and scientific merit of the technical papers produced by each team. The final evaluation is performed by judges interacting on a screen-equipped device in a closed session.
“There are two criteria for evaluation: rating and conversation duration. A team with a rating of at least 4.0 out of 5.0 and a conversation length of 20 minutes gets the final prize. Within the final evaluation period, our team had been ranked in the top three according to users' average ratings with our latest version of the bot reaching a 35:04-minute mean conversation length.”
Being able to successfully conduct very long conversations is part of what sets NAM apart from other bots.
“On average, our bot's conversation time doubles other bots' time. For some users, our bot reaches 56 minutes of chat duration, compared to about 10 to 20 minutes typically,” Xu says.
Another differentiator: NAM is self-learnable through interaction with users. The Stevens bot asks clarifying or follow-up questions to clear up confusion or uncertainty during the conversation. NAM is also trained to provide explanations when making suggestions.
The product team in this final stage includes first-year master’s students João Luís Lins Rodrigues Cruz, Sai Nikhil Reddy Maligireddy, Abhijeet Gusain and Yeshwanth Reddy Peddamallu. The team is also helped by Xu’s postdoc Abdul Rafae Khan and supported by all her Ph.D. students.
No matter the outcome of the final round of competition, Xu is proud of the team’s accomplishment.
“Most of our team members at this point are fresh students without prior natural language processing backgrounds, and they became chatbot experts through the competition,” she says. “We realized our unique social bot based on imagination, innovation and technology, and we are happy to be in the final.”
Alexa customers can interact with the university socialbots, including NAM, by saying "Alexa, let’s discuss" on Amazon Echo, Fire TV or tablet devices, or by using the Alexa app on their phone or computer.